JP2009022017A - Image processing apparatus and blur-detecting method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately detect blurs in an image even in a high-resolution image, while reducing memory capacity to be used. <P>SOLUTION: A printer 100 acquires a horizontal DCT coefficient group and a vertical DCT coefficient group for each block from JPEG data, selects an edge pattern approximate to a luminance change represented with these coefficient groups from a predetermined table and records a pattern number thereof in a RAM 170. The printer 100 decides whether the luminance changes in blocks horizontally and vertically adjacent to each other continue, based on pattern numbers of blocks recorded in the RAM 170 and links edge patterns, when the luminance changes continue, to accumulate the width of the blur that exists over the blocks. Based on such a cumulative value, the presence or the absence of blurs of an image is determined. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for detecting blur of an image.

  In recent years, digital still cameras have become widespread, and the capacity of memory cards used for them has increased. Therefore, the opportunity for a general user to take a large number of images is increasing. Since a digital still camera does not require a film fee and can easily shoot, it is often performed without being aware of subject blurring or camera shake. Therefore, when a photographed image is to be printed by a printer, there are relatively many images in which the image is blurred due to subject blurring or camera shake, and it is necessary in advance to select a normal image. .

  The operation of selecting a normal image from a large number of images is a very complicated operation. Therefore, in order to cope with such a problem, there is a demand for a technique for automatically removing a blurred image from a printing target before the user prints the image. In relation to such blur detection technology, Patent Document 1 below divides a captured image into a plurality of blocks each having 8 pixels × 8 pixels, performs frequency analysis on each block subjected to discrete cosine transform, A technique for determining the sharpness of an image (in other words, the degree of blurring) is described for each block.

Japanese Patent Laid-Open No. 4-170872 JP-A-10-271516 JP 2002-51178 A

  However, since the resolution of an image photographed by a recent digital still camera is a high resolution of several million to 10 million pixels, as described in Patent Document 1, it is within a block composed of only 8 pixels × 8 pixels. It was difficult to determine blur and sharpness only. Specifically, when an image of 6 million pixels (about 3000 pixels wide × 2000 pixels high) is printed on an L-size printing paper (about 130 mm wide × 90 mm long), 8 pixels wide is about 0.00 mm. It corresponds to a width of 35 mm, and it is difficult to objectively determine whether an image is sharp or blurred only within such a narrow width. That is, it has become difficult to directly apply the technique described in Patent Document 1 to high-resolution images taken in recent years.

  Further, when trying to analyze blur of an image having a high resolution of several million pixels or more by the technique described in Patent Document 1, the amount of calculation becomes enormous, and the amount of memory used increases accordingly. Such a problem is a problem that cannot be overlooked particularly in the case of detecting blur in a small device with limited CPU power or memory capacity, such as a printer, a digital camera, or a photo viewer.

  In view of the various problems described above, the problem to be solved by the present invention is to detect blurring of an image with high accuracy even in a high-resolution image while reducing the memory capacity to be used.

Based on the above problems, the image processing apparatus of the present invention is configured as follows. That is,
An image processing apparatus for detecting blur of an image configured as a set of blocks composed of a plurality of pixels,
An image data input unit for inputting image data in which coefficients obtained by conversion from a spatial domain to a frequency domain performed in units of the blocks are recorded for each block; and
A plurality of types of basic edge patterns in which representative gradient shapes of pixel value changes in the block are represented by the coefficients, and pattern numbers uniquely assigned to the basic edge patterns are stored in association with each other. An edge pattern storage unit;
A coefficient extraction unit that extracts a coefficient group representing a frequency component in a predetermined direction from each block of the input image data;
A pattern matching unit that selects a basic edge pattern that approximates a gradient shape represented by the extracted coefficient group from the edge pattern storage unit;
A pattern number storage unit that stores the pattern number corresponding to the selected basic edge pattern in association with the block;
A blur width detecting unit that refers to a pattern number stored in the pattern number storage unit and detects a range in which the gradient directions of the basic edge pattern coincide with each other in a series of blocks in the image as a blur width;
And a blur determination unit that determines blur of an image represented by the image data based on the detected blur width.

  In the image processing apparatus of the present invention, a coefficient group representing a frequency component in a predetermined direction is extracted for each block constituting image data, and a basic edge pattern approximated to the shape of the gradient of the pixel value represented by the coefficient group is defined as an edge pattern. The pattern number is selected from the storage unit and stored in the pattern number storage unit. Then, based on the pattern number, the connection state of the basic edge patterns is determined along a series of blocks, and a range in which the gradient directions of the basic edge patterns match is detected as a blur width. Therefore, even when the blur width exceeds the size of one block, it can be determined whether or not the image is blurry with high accuracy.

  Further, as described above, in the present invention, in determining whether the gradient directions of the basic edge patterns between adjacent blocks coincide with each other, instead of using the coefficient recorded in the image data, the gradient represented by that coefficient is used. The pattern number of the basic edge pattern that approximates the shape is used. Therefore, the processing time is greatly shortened, and the memory capacity to be used can be significantly reduced. For example, if the pattern number associated with a block is the number of a basic edge pattern representing a downward slope, and the pattern number associated with the adjacent block is also the number of a basic edge pattern representing a downward slope Therefore, it can be determined that the gradient directions coincide between these blocks. Note that “the direction of the gradient is the same” does not mean that the gradient angle is the same. For example, if the signs of the slopes match, it can be said that the directions of the slopes match.

  In the present application, image data in which coefficients obtained by conversion from the spatial domain to the frequency domain are recorded for each block is input. As such image data, for example, image data in JPEG format is used. There is.

In the image processing apparatus having the above configuration,
The edge pattern storage unit further stores a gradient width represented by each basic edge pattern in association with the pattern number,
The blur width detection unit reads the gradient width associated with the pattern number of each block existing in a range where the gradient directions of the basic edge pattern coincide with each other from the edge pattern storage unit, and calculates the gradient width. The blur width may be calculated by addition.

  With such a configuration, since the gradient width represented by each basic edge pattern is stored in advance in the edge pattern storage unit, the blur width can be calculated at high speed.

In the image processing apparatus having the above configuration,
The blur width detection unit moves a block of interest in the image data along the series of blocks, and is associated with the block of interest and an adjacent block to which the block of interest is moved, respectively. Referring to the pattern number, it is determined whether or not the direction of the gradient of the basic edge pattern of the block of interest and the adjacent block matches, and if the direction of the gradient matches, it corresponds to the pattern number of the block of interest The blur width may be calculated by accumulatively adding the attached gradient widths.

  With such a configuration, the block of interest is sequentially moved in the image, and the blur width is calculated while accumulating the gradient width, so that the memory usage required for calculating the blur width can be reduced. it can.

In the image processing apparatus having the above configuration,
In the image data, the blocks are arranged in a horizontal direction and a vertical direction,
The coefficient extraction unit extracts, as the coefficient group, a first coefficient group representing a horizontal frequency component and a second coefficient group representing a vertical frequency component,
The pattern matching unit selects the basic edge pattern for each of the first coefficient group and the second coefficient group,
The pattern number storage unit stores the pattern numbers corresponding to the selected two basic edge patterns in association with the blocks, respectively.
The blur width detection unit calculates the blur width in the horizontal direction and the vertical direction,
The blur determination unit may determine blur of an image represented by the image data based on a larger blur width among the blur widths calculated for the horizontal direction and the vertical direction, respectively.

  According to such a configuration, the horizontal blur width and the vertical blur width can be calculated for one block. Therefore, if the larger blur width is used, it is possible to calculate the image with higher accuracy. It becomes possible to determine the blur of the image.

In the image processing apparatus having the above configuration,
The blur width detecting unit stores the gradient width cumulatively added in the horizontal direction in a horizontal cumulative buffer that holds one block, and holds the gradient width cumulatively added in the vertical direction for one block. When the block of interest is moved and stored in the vertical accumulation buffer, the horizontal and vertical gradient widths accumulated in the block before the movement may be acquired by referring to these buffers.

  According to such a configuration, by preparing the minimum necessary buffers such as a horizontal accumulation buffer that holds an accumulation value for one block and a vertical accumulation buffer that holds an accumulation value for one row of blocks, The gradient width can be accumulated for each vertical direction. As a result, memory usage can be reduced.

In the image processing apparatus having the above configuration,
The pattern number storage unit has a storage area capable of storing the pattern number for a first row including the target block and a second row including a block adjacent to the lower side of the target block, When the block of interest moves to the right end of the row to which the block of interest belongs, the pattern number stored in the second and subsequent rows is moved up to the previous row and stored, so that the next selected block The pattern number may be stored in an empty area of the pattern number storage unit.

  According to such a configuration, it is only necessary to store the pattern number only for the row including the block adjacent to the target block, not the pattern number of the entire image, so that the memory capacity to be used can be greatly reduced. Become.

In the image processing apparatus having the above configuration,
In the image data, the block is recorded every n (n is an integer of 1 or more) rows,
The pattern number storage unit may prepare a storage area capable of storing pattern numbers for the block (n + 1) rows.

  According to such a configuration, the minimum necessary pattern number can be stored according to the number of unit rows in which blocks are recorded in the image data, so that the memory capacity can be saved. If the image data is based on the JPEG standard, the block is recorded in units of one or two lines, so the pattern number storage unit can store pattern numbers for two or three lines. An area can be prepared. It can be determined by analyzing the header information of the JPEG data that the number of rows is recorded as a block.

In the image processing apparatus having various configurations described above,
The blur determination unit
A block blur determination unit that determines the presence or absence of blur for each block based on the blur width and a predetermined threshold;
For all blocks included in the predetermined window area in the image data, the presence / absence of blur by the block blur determination means is totaled, and whether or not the window area is blurred based on the total result and a predetermined threshold value A window blur determination unit for determining
As a result of moving the window area in the image data, if there is at least one window area that is determined not to be blurred, it is determined that the entire image represented by the image data is not blurred. It is good also as what has an image blur judgment part to perform.

  According to such a configuration, it is possible to determine the presence or absence of blur for each window area composed of a plurality of blocks, so it is possible to determine whether or not the image is blurred with an accuracy closer to human sensitivity. It becomes possible. The predetermined window area may be, for example, an area of 1 cm × 1 cm or 2 cm × 2 cm assuming that image data is printed on an L-size printing paper. This is because if it is determined that the window area of such a size is not blurred in any part of the image, it can be determined that the area is in focus.

In the image processing apparatus having the above configuration,
The window area is further partitioned by a plurality of sub-windows, each of which is larger than the size of the block,
The block blur determination unit totalizes the presence / absence of blur for each block for each sub-window including the block,
The window blur determination unit further counts the total value of the sub-windows, and further totals all the sub-windows included in the window area, thereby calculating the presence / absence of the blur for all blocks included in the window area. It may be done.

  According to such a configuration, the presence / absence of blurring may be tabulated in a stepwise manner from the block to the subwindow and from the subwindow to the window area, so that the presence / absence of blurring can be efficiently tabulated.

In the image processing apparatus having the above configuration,
The image blur determination unit may move the window area within the image data by a movement distance in units of sub-windows.

  With such a configuration, since the window area moves by the movement distance in units of sub-windows, it is possible to easily add up the presence or absence of blur in the window area. In addition, it is possible to detect a blurred window area with higher accuracy than when moving in window area units.

In the image processing apparatus having the above configuration,
Furthermore, a total value storage unit that stores a total value of the presence or absence of blur for each of the sub-windows totaled by the block blur determination unit,
The total value storage unit has a storage area for storing the total value as many as the number of rows of subwindows included in the window area, and when the window area is moved downward in the image data The total value stored in each row of the total value storage unit is stored in the previous row by the number of the moved lines, and the empty region of the total value storage unit generated by the advance processing The total value for each of the sub-windows that are included in the moved window area may be stored.

  According to such a configuration, it is only necessary to prepare as the total value storage unit a memory capacity necessary for totalizing the presence / absence of blur in the window area, so that the memory capacity to be used can be reduced.

In the image processing apparatus having the above configuration,
Furthermore, it is good also as a thing provided with the presentation part which shows a user the image determined not to be blurred by the blur determination part.

  According to such a configuration, the user can confirm only images that have been successfully captured. As a method of presenting to a user, for example, an image can be displayed on a display device. In addition, a list of images determined not to be blurred may be printed and presented to the user.

In the image processing apparatus having the above configuration,
Furthermore, a printing unit that prints an image selected by the user from the presented images may be provided.

  According to such a configuration, the user can easily print a desired image from images that have been successfully shot without being aware of an image that is blurred due to camera shake or subject blur. .

  In addition to the configuration as the image processing apparatus described above, the present invention can also be configured as a blur detection method in which a computer detects blur of an image and a computer program for detecting blur in an image. Such a computer program may be recorded on a computer-readable recording medium. As the recording medium, for example, various media such as a flexible disk, a CD-ROM, a DVD-ROM, a magneto-optical disk, a memory card, and a hard disk can be used.

Hereinafter, in order to further clarify the operations and effects of the present invention described above, embodiments of the present invention will be described based on examples in the following order.
A. Printer hardware configuration:
B. Printer functional configuration:
C. Printing process:
D. Blur judgment processing:
E. Edge pattern matching process:
F. Edge connection processing:
G. Block blur judgment processing:
H. Window blur determination processing:
I. effect:

A. Printer hardware configuration:
FIG. 1 is an explanatory diagram showing the appearance of a printer 100 as an embodiment of the image processing apparatus of the present application. The printer 100 is a so-called multi-function printer, and is connected to devices such as a scanner 110 for optically reading an image, a memory card slot 120 for inserting a memory card MC on which image data is recorded, and a digital camera. A USB interface 130 and the like are provided. The printer 100 can print an image captured by the scanner 110, an image read from the memory card MC, and an image read from the digital camera via the USB interface 130 on the printing paper P. It is also possible to print an image input from a personal computer (not shown) connected by a printer cable or USB cable.

  The printer 100 includes an operation panel 140 for performing various operations related to printing. A liquid crystal display 145 is provided at the center of the operation panel 140. The liquid crystal display 145 is used to display an image read from a memory card MC, a digital camera, or the like, or to display a GUI (graphical user interface) when using various functions of the printer 100.

  The printer 100 eliminates a blurred image (hereinafter referred to as “blurred image”) from a plurality of image data input from a memory card MC, a digital camera, or the like, and an image that is in focus at one location (hereinafter referred to as “blurred image”). , “Focused image”), and a function for displaying on the liquid crystal display 145. The user can print only an image suitable for printing by selecting a desired image from the images displayed on the liquid crystal display 145 in this way. Hereinafter, the configuration and processing of the printer 100 for realizing the function of eliminating the blurred image will be described in detail.

  FIG. 2 is an explanatory diagram showing the internal configuration of the printer 100. As shown in the figure, the printer 100 has a carriage 210 on which an ink cartridge 212 is mounted, a carriage motor 220 that drives the carriage 210 in the main scanning direction, and a printing paper P in the sub-scanning direction as a mechanism for printing on the printing paper P. A paper feed motor 230 and the like are provided.

  The carriage 210 includes a total of six types of ink heads 211 corresponding to the inks representing cyan, magenta, yellow, black, light cyan, and light magenta. An ink cartridge 212 containing these inks is mounted on the carriage 210, and the ink supplied from the ink cartridge 212 to the ink head 211 is ejected onto the printing paper P by driving a piezo element (not shown). The

  The carriage 210 is movably held by a sliding shaft 280 installed in parallel with the axial direction of the platen 270. The carriage motor 220 rotates the drive belt 260 in response to a command from the control unit 150 to reciprocate the carriage 210 in parallel with the axial direction of the platen 270, that is, in the main scanning direction. The paper feed motor 230 rotates the platen 270 to convey the printing paper P perpendicular to the axial direction of the platen 270. That is, the paper feed motor 230 can relatively move the carriage 210 in the sub-scanning direction.

  The printer 100 includes a control unit 150 for controlling operations of the ink head 211, the carriage motor 220, and the paper feed motor 230 described above. The control unit 150 is connected to the scanner 110, the memory card slot 120, the USB interface 130, the operation panel 140, and the liquid crystal display 145 shown in FIG.

  The control unit 150 includes a CPU 160, a RAM 170, and a ROM 180. The ROM 180 stores a control program for controlling the operation of the printer 100, and further stores an edge pattern table 181 and a normalized luminance difference table 182 that are used in various processes described later.

  The CPU 160 operates as each functional unit (161 to 166) shown in the figure by expanding and executing the control program stored in the ROM 180 in the RAM 170. The RAM 170 includes a work area for executing the control program, a pattern number buffer 171 and a focused block number buffer 172 (corresponding to the total value storage unit of the present application), a vertical accumulation buffer 173, and a horizontal accumulation buffer 174 shown in the figure. A storage area called "," is secured.

B. Printer functional configuration:
The control unit 150 includes an image data input unit 161, a coefficient extraction unit 162, a pattern matching unit 163, an edge connection unit 164, a block blur determination unit 165, and a window blur determination unit 166 as functional units realized by the CPU 160. . Hereinafter, the function of these functional units will be briefly described (for details of the function, refer to various processing contents described later).

  The image data input unit 161 is a functional unit that inputs JPEG format image data (hereinafter referred to as “JPEG data”) from the memory card MC or the digital camera via the memory card slot 120 or the USB interface 130. In JPEG data, blocks called MCU (Minimum Coded Unit) of 8 pixels × 8 pixels are arranged in a row in the horizontal direction and the vertical direction (that is, for each row). The image data in each block includes (1) color conversion of pixel values from the RGB color space to the YCbCr color space, (2) DCT (Discrete Cosine Transform) conversion from the spatial domain to the frequency domain, (3) The amount of data is compressed by sequentially performing quantization for reducing the amount of information and (4) Huffman coding, which is a kind of entropy coding.

  3 and 4 are explanatory diagrams showing the order in which the data of each block is recorded in the JPEG data. FIG. 3 shows a state in which each block is recorded for each row, and FIG. 4 shows a state in which each block is recorded for every two rows. When blocks are stored every two rows, as shown in FIG. 4, each block has an alphabet “Z” sequentially from the upper left of the image in the order of right, lower left, right, upper right. It is recorded as if it were written. In the header information of JPEG data, information indicating whether the blocks are recorded in any order is recorded. In addition, some JPEG data has data of each block recorded in another recording order. However, in this embodiment, only JPEG data in which data is recorded by the two types of recording orders described above. Shall be handled. In this embodiment, the continuous direction of the blocks is assumed to be the horizontal direction and the vertical direction, and the continuous direction and the recording order of the blocks are different.

  The coefficient extraction unit 162 performs inverse Huffman coding and inverse quantization on the above-described JPEG data for each block, and acquires DCT coefficients. The coefficient extraction unit 162 further includes a first coefficient group representing a horizontal AC frequency component and a second coefficient group representing a vertical AC frequency component from the 8 × 8 DCT coefficients thus obtained. And extract.

  FIG. 5 is an explanatory diagram showing DCT coefficients extracted by the coefficient extraction unit 162. As shown in the figure, a total of 64 DCT coefficients from F00 to F77 are obtained from each block when the inverse quantization is performed. Among these, the coefficient F00 present at the upper left is called a DC component, and the other coefficients are called AC components. Among these, the coefficient extraction unit 162 extracts the coefficients F01 to F07, which are AC components only in the horizontal direction, as the first coefficient group, and the coefficients F10 to F70, which are AC components only in the vertical direction, as the second coefficient group. Extract.

  The pattern collating unit 163 collates the edge pattern table 181 stored in the ROM 180 with the first coefficient group and the second coefficient group extracted by the coefficient extracting unit 162, and thereby compares the first coefficient group and the second coefficient group. This is a functional unit that selects a basic edge pattern having a shape approximate to the luminance change represented by the coefficient group from the edge pattern table 181. In this embodiment, the “edge pattern” refers to a pattern that represents the shape of the gradient of luminance change in a block. The detailed configuration of the edge pattern table 181 will be described later. In this edge pattern table 181, DCT coefficient group values representing the gradient shapes and pattern numbers of 64 types of basic edge patterns are recorded in association with each other. ing. In the lower part of FIG. 5, an example of selecting a basic edge pattern approximated to the horizontal luminance change represented by the first coefficient group and a basic edge pattern approximated to the vertical luminance change represented by the second coefficient group. Is shown. When the basic edge pattern is selected in this way, the pattern matching unit 163 records the pattern number of the selected basic edge pattern in the pattern number buffer 171 in the RAM 170 for each block.

  The edge connecting unit 164 is a functional unit that performs processing for connecting the width of the blurred portion (hereinafter referred to as “blurred width”) when a blurred outline portion exists across a plurality of blocks. The edge connecting portion 164 corresponds to the “blurring width detecting portion” of the present application.

  FIG. 6 is an explanatory diagram showing a concept of connecting blur widths of a block of interest and a block adjacent thereto. In this embodiment, as shown in the upper part of the figure, based on the shape of the basic edge pattern selected for the target block and the basic edge pattern selected for the right and lower adjacent blocks, the luminance change of these blocks It is determined whether or not the gradients of are continuous in the same direction. If it is continuous, the blur width and luminance difference of the adjacent blocks on the upper and left sides are sequentially accumulated, and if not continuous, it is determined that the target block is the end of the luminance change, Processing to determine the blur width and brightness difference is performed. That is, in this embodiment, when the blur width of the image exceeds the size of one block (8 × 8 pixels), the blur width is cumulatively added, and when the block reaches the end of blur. Based on the cumulative addition value (hereinafter, referred to as “cumulative blur width CW”), it is determined whether the block is in focus or blurry. As shown in the lower part of the figure, when the blur width is cumulatively added for the block of interest, the value is referred to when calculating the blur width of the right and lower blocks. Note that the edge linking unit 164 has a function of accumulating not only the blur width but also the luminance difference in the block, and improving the determination accuracy of the presence / absence of blur based on this value.

  The block blur determination unit 165 is a functional unit that determines, for each block, whether the block is blurry. Specifically, when the cumulative blur width CW having a larger value among the horizontal and vertical cumulative blur widths CW determined by the edge linking unit 164 is adopted and this blur width exceeds a predetermined threshold value, The block is determined to be blurred.

  The window blur determination unit 166 is a functional unit that determines whether or not the window area of the window area configured by a plurality of blocks is blurred. In the following description, a focused window area is referred to as a “focused window”, and a blurred window area is referred to as a “blurred window”.

  FIG. 7 is an explanatory diagram showing the concept of the window area. In this embodiment, the window area represents an area having a size of 1 cm × 1 cm, which is assumed to be the minimum focusing area on the L-size printing paper. The window blur determination unit 166 determines that the image is a focused image if the window region is in focus at any one position in the L-size printing paper. The L-size printing paper has a size of about 13 cm × length of about 9 cm. When an image of 6 million pixels (approximately 3000 pixels wide × approximately 2000 pixels long) is printed on this sheet, the size of one block of 8 pixels × 8 pixels is about 0.35 mm × 0.35 mm. Therefore, there are about 28 × 28 blocks in one window area.

  As shown in the lower part of FIG. 7, the window area is divided into areas called 4 × 4 sub-window areas. The window blur determination unit 166 determines which window region in the image is in focus by moving the position of the window region in units of the sub-window region. In the example described above, there are 28 × 28 blocks in one window area, and therefore there are 7 × 7 blocks in this sub-window area.

  In this embodiment, as described above, the number of blocks and the window size are defined on the assumption that an image of 6 million pixels is printed on an L-size printing paper. It goes without saying that the values of these parameters are appropriately changed according to the size and the resolution of the image.

C. Printing process:
FIG. 8 is a flowchart of print processing executed by the CPU 160 of the printer 100. This printing process is a process for printing by inputting image data from the memory card MC or the like.

  When this printing process is executed in accordance with a predetermined operation by the user using the operation panel 140, first, the CPU 160 uses the image data input unit 161 to perform JPEG from the memory card MC inserted in the memory card slot 120. Data is input (step S10). Here, JPEG data is input from the memory card MC, but may be input from a digital camera or a computer connected via the USB interface 130.

  When the JPEG data is input, the CPU 160 analyzes the header information of the input JPEG data and secures a buffer area necessary for the subsequent processing in the RAM 170 (step S20). Specifically, when the block is recorded in the unit of one line as shown in FIG. 3 in the JPEG data, the CPU 160 records the pattern number buffer 171 that can record the pattern number of the basic edge pattern for two lines of the block. Is secured in the RAM 170. When the block is recorded in units of two lines as shown in FIG. 4, a pattern number buffer 171 capable of recording pattern numbers for three lines of blocks is secured. That is, if a block is recorded every n (n is an integer of 1 or more) rows in the JPEG data, the CPU 160 secures a pattern number buffer 171 that can record a pattern number of blocks of (n + 1) rows. It is. Further, the CPU 160, in addition to the pattern number buffer 171, a vertical accumulation buffer 173 that records the cumulative value of blur width in the vertical direction for one block, and a horizontal accumulation that records the cumulative value of blur width in the horizontal direction for one block. A buffer 174 and an in-focus block number buffer 172 for recording the in-focus block number for four rows of the sub-window are secured in the RAM 170. The role and usage of these buffers will be described as appropriate below.

  Next, the CPU 160 performs a blur determination process on the JPEG data input in step S10 while using the various buffers secured in step S20 (step S30). Details of the blur determination process will be described later.

  When the blur determination process is completed for one piece of JPEG data, the CPU 160 inputs all the JPEG data in the memory card MC and determines whether the blur determination process is completed for these (step S40). If it is determined that all the JPEG data has not been input through this process (step S40: No), the process returns to step S10, and the next JPEG data is input.

  If it is determined in step S40 that the blur determination process has been completed for all the JPEG data (step S40: Yes), the CPU 160 determines the image determined not to be blurred in step S30, that is, the focused image. Then, a list is displayed on the liquid crystal display 145 (step S50).

  When the list of focused images is displayed on the liquid crystal display 145, the CPU 160 accepts selection of an image to be printed from the user via the operation panel 140 (step S60). Then, the selected image is printed by controlling the ink head 211, the paper feed motor 230, the carriage motor 220, and the like (step S70).

  In the printing process described above, all the JPEG data in the memory card MC is input. However, when a plurality of folders are generated in the memory card MC, only the folder designated by the user. JPEG data may be input.

D. Blur judgment processing:
FIG. 9 is a flowchart of the blur determination process executed in step S30 of the printing process shown in FIG. When this process is started, first, the CPU 160 moves blocks to be subjected to blurring determination (hereinafter referred to as “target block”) along the series (step S100). The first destination is the block that exists at the top left of the JPEG data. The block of interest moves to the right every time the process of step S100 is executed, and when it reaches the right end, it moves to the left end of the next row. The movement order of the block of interest is an order independent of the block storage order shown in FIGS.

  Next, the CPU 160 determines whether the blocks adjacent to the right side and the lower side of the current block of interest have already been read (step S110). This is because, as shown in FIG. 6, the blur width and luminance difference of the central block (target block) are obtained based on the data of adjacent blocks. If the blocks are read in accordance with the storage order of the blocks in FIG. 3 or FIG. 4, if the reading of the right and lower blocks is completed, it can be said that the reading of the blocks adjacent to the upper and left sides is also completed.

  If the reading of the blocks adjacent to the right side and the lower side of the target block has not been completed (step S110: No), the CPU 160 reads one block data from the JPEG data according to the order shown in FIG. 3 or FIG. Step S120). Since the header information of JPEG data records information indicating in which order the blocks are stored, the CPU 160 can determine the reading order of the blocks by analyzing the information.

  When one block is read, the CPU 160 performs inverse Huffman coding and inverse quantization on the block data, and then uses the coefficient extraction unit 162 to select a horizontal direction from the 8 × 8 coefficient group. The first coefficient group and the second coefficient group in the vertical direction are extracted (step S130). Specifically, the DCT coefficients from “F01” to “F07” shown in FIG. 5 are extracted as the first coefficient group, and the DCT coefficients from “F10” to “F70” are extracted as the second coefficient group. To do.

  Next, the CPU 160 executes edge pattern matching processing using the pattern matching unit 163 (step S140). In this process, a basic edge pattern that approximates the horizontal and vertical gradient shapes represented by the first coefficient group and the second coefficient group extracted in step S130 is selected from the edge pattern table 181. It is a process (detailed processing content will be described later). If an approximate basic edge pattern is selected by such processing, the pattern number assigned to the basic edge pattern is stored in the pattern number buffer 171 secured in the RAM 170.

  FIG. 10 is an explanatory diagram showing an example of the pattern number buffer 171 secured when a block is recorded in JPEG data in units of one line. In this figure, in addition to the pattern number buffer 171, examples of the vertical accumulation buffer 173 and the horizontal accumulation buffer 174 are also shown. As shown in the figure, when blocks are stored in JPEG data in units of one line, the pattern number buffer 171 can store the pattern numbers in the horizontal direction and the vertical direction for two blocks. In the figure, an example is shown in which horizontal pattern numbers (A2, A4, B1, C6,...) Are recorded up to a block adjacent to the lower side of the target block. Although not shown in the figure, in the edge pattern matching process executed in step S140, the luminance difference DY in the block is also calculated, so the luminance difference DY is also recorded in the pattern number buffer 171 together with the pattern number. The

  When blocks are stored in JPEG data in units of one line, it is assumed that the block of interest moves from the left to the right in the pattern number buffer 171 shown in FIG. In other words, the target block sequentially moves from the left to the right in the first row in the pattern number buffer 171. When the movement to the right end is completed, the movement is again made to the left end of the first row. At this time, the pattern number recorded in the second line of the pattern number buffer 171 is moved up by one line and recorded, and the pattern number is cleared for the second line. Then, the pattern numbers acquired by the next edge pattern matching process are sequentially recorded on the second line. In this way, when the block of interest moves to the right end, if the next pattern number is recorded by moving up the row of the pattern number buffer 171, the block of interest moves relatively to the lower side of the image.

  When blocks are stored in units of one row for JPEG data, it is obvious from FIG. 10 to determine whether adjacent blocks on the right side and lower side of the target block have already been read. Thus, it is sufficient that at least the pattern numbers for two rows of blocks are recorded. Therefore, in this embodiment, when a block is stored in units of one line for JPEG data, a pattern number buffer 171 capable of recording pattern numbers for two lines of blocks is prepared. By doing so, it is not necessary to store the pattern numbers for all the blocks included in the JPEG data, and the necessary memory capacity can be reduced.

  FIG. 11 is an explanatory diagram showing an example of the pattern number buffer 171 secured when blocks are recorded in JPEG data in units of two lines. As shown in the figure, in this case, the pattern number buffer 171 is provided with a storage area for pattern numbers corresponding to three rows of blocks.

  When blocks are stored in JPEG data in units of two lines, if the block of interest is located on the first row in the image, in step S100, as shown in FIG. The first line in the pattern number buffer 171 is moved to the right. When the movement to the right end is completed, the movement is again made to the left end of the first row. At this time, the pattern number recorded in the second line of the pattern number buffer is moved up by one line and recorded, and the pattern number is cleared for the second line. Then, as shown in FIG. 11B, using the second and third lines, the pattern numbers acquired by the edge pattern matching process are recorded every two lines according to the arrangement order shown in FIG. Go.

  To determine whether or not adjacent blocks on the right and bottom sides of the target block have already been read when JPEG data is stored in units of two rows, see FIG. 11B. As is clear, at least a pattern number for three rows of blocks is required. FIG. 11B shows an example in which the right and lower blocks of the target block (pattern number: C2 in the figure) in the first row in the pattern number buffer 171 are recorded. This is because the lower block (pattern number: A7 in the figure) is not read unless the second block ahead of the current row is read according to the storage order shown in FIG. Therefore, in this embodiment, when a block is stored in units of two rows for JPEG data, a pattern number buffer 171 capable of storing pattern numbers for three rows of blocks is prepared.

  FIGS. 10 and 11 show a vertical cumulative buffer 173 for storing the vertical cumulative blur width CW and cumulative luminance difference ADY for one block, and a horizontal cumulative blur width CW for one block. The horizontal cumulative buffer 174 for storing the cumulative luminance difference ADY is shown. A method of using these buffers will be described later.

  Here, the description returns to FIG. When the pattern number of the block read in step S120 is stored in the pattern number buffer 171 by the edge pattern matching process in step S140, the CPU 160 executes the process in step S110 again, so that the right side of the current block of interest And it is determined whether or not a block adjacent to the lower side has been read. Even if it is determined that these blocks have not been read by such processing, the processing from step S120 to step S140 is repeated until the adjacent blocks on the right side and the lower side are read, and the pattern number buffer The pattern number is stored in 171.

  If it is determined in step S110 that adjacent blocks on the right side and the lower side of the target block have been read (step S110: Yes), the CPU 160 uses the pattern number buffer 171, the vertical accumulation buffer 173, and the horizontal accumulation buffer 174. With reference to the edge connecting unit 164, an edge connecting process is executed (step S150). Although details of the edge linking process will be described later, in this edge linking process, the blur widths of adjacent blocks are cumulatively added, and the blur width of the target block is calculated.

  When the blur width of the block of interest is calculated by the edge linking process, the CPU 160 then executes the block blur determination process using the block blur determination unit 165 (step S160). This process is a process for determining whether or not the block of interest is blurred based on the blur width calculated by the edge connection process. Although details of this process will be described later, as the execution result of this process, whether the target block is a “blurred block” or a “focus block” is acquired as a determination result.

  When the block blur determination process determines whether the target block is a “blurred block” or “focused block”, the CPU 160 determines in which sub-window (see FIG. 7) the current target block is located. And the cumulative number of in-focus blocks is added for each sub-window (step S170). FIG. 7 shows an example in which a total of 49 blocks of 7 × 7 are present in one sub-window. In this case, the cumulative added value by such processing is “49” at the maximum. The cumulative added value of the focused block is recorded in the focused block number buffer 172 secured in the RAM 170.

  FIG. 12 is an explanatory diagram showing an example of the focused block number buffer 172. In this embodiment, in order to finally determine whether or not the window area corresponding to 4 × 4 sub-windows is blurred, the in-focus block number buffer 172 has a storage area for four rows of sub-windows as shown in the figure. Prepare. In step S170, the number of focused blocks is cumulatively added to the storage area corresponding to the subwindow to which the current block of interest belongs. The accumulated values are recorded in the order of “a”, “b”, “c”,... Shown in the figure from the fourth line of the focused block number buffer 172. When the subwindow to which the current block of interest belongs moves to the right end of the focused block number buffer 172, the value in the fourth row of the focused block number buffer 172 is moved up to the third row, Lines are also moved up line by line. Then, the buffer area in the fourth row is cleared, and the next focused block number is recorded in this area.

  Returning to FIG. After accumulating and adding the number of in-focus blocks for each sub-window in step S170, CPU 160 determines whether or not the total number of in-focus blocks has been completed for any one sub-window (step S180). If it is determined that the aggregation for one sub-window has not been completed (step S180: No), the CPU 160 returns the process to step S100, moves the block of interest, and continues to “blurred” for the next block. It is determined whether it is a “block” or a “focus block”. On the other hand, if it is determined that the aggregation for one sub-window has been completed, the CPU 160 performs window blur determination processing using the window blur determination unit 166 for the entire window region to which the sub-window belongs (step 190). In this window blur determination process, the window area is sequentially moved in the horizontal direction in the focused block number buffer 172 shown in FIG. 12, and the window area is determined based on the total number of focused blocks in the window area. Judge whether it is blurry or not. Details of this processing will be described later.

  When the window blur determination process ends in step S190, the CPU 160 determines whether or not the current window area is a focused window based on the determination result of the determination process (step S200). If the result of this determination is that the window is an in-focus window (step S200: Yes), it is determined that the image represented by the currently input JPEG data is in focus (step S210), and the series described above. The blur determination process is terminated. That is, when it is determined that any window area in the image is a focused window, it is determined that the JPEG data input in step S10 in FIG. 8 is a focused image. In this way, if any window region in the image is a focused window, it is not necessary to determine the presence or absence of blurring for all the regions in the image, so that the processing time can be shortened.

  On the other hand, when it is determined that the window area subjected to the window blur determination process in step S200 is a blur window (step S200: No), the CPU 160 subsequently determines that the current block of interest is It is determined whether it is the last block in the lower right corner of the image (step S220). As a result, if it is determined that the current block of interest is the final block (step S220: Yes), it can be determined that no window area is a focused window, and therefore it is represented by the currently input JPEG data. It is determined that the image is a blurred image (step S230), and the above-described series of blur determination processing ends. On the other hand, if the current block of interest is not the final block (step S220: No), the CPU 160 returns the process to step S100, and repeatedly executes the various processes described above for the block at the next position.

E. Edge pattern matching process:
FIG. 13 is a detailed flowchart of the edge pattern matching process executed in step S140 of the blur determination process shown in FIG. This process is a process for selecting a basic edge pattern that approximates the horizontal and vertical gradient shapes of each block from the edge pattern table 181 and calculating a luminance difference within the block.

  When this process is executed, first, the CPU 160 acquires the first coefficient group and the second coefficient group (see FIG. 5) extracted in step S130 of the blur determination process (step S300). In the edge pattern matching processing, the same processing is executed for each of the first coefficient group and the second coefficient group. Accordingly, in the following, the first coefficient representing the DCT coefficient in the horizontal direction will be representatively represented. Processing for the group will be described.

  When the CPU 160 acquires the first coefficient group in step S300, the CPU 160 obtains a sum of absolute values of the coefficients F01 to F07 constituting the first coefficient group, and this obtains a predetermined flat threshold (for example, 80). It is determined whether or not it exceeds (step S310). The sum S of absolute values of the first coefficient group can be obtained by the following equation (1).

S = Σ | F0i | (i = 1-7) (1)

  If it is determined in step S310 that the sum S of absolute values of the first coefficient group is equal to or less than a predetermined flat threshold value (step S310: No), the luminance change in the horizontal direction of the block is flat. Can be considered. Accordingly, the pattern number of the edge pattern of the block is determined as “0” indicating that it is a flat edge pattern (step S320).

  If it is determined in step S310 that the sum S of absolute values of the first coefficient group exceeds a predetermined flat threshold (step S310: Yes), it is determined that there is some luminance change in the horizontal direction of the block. it can. Therefore, first, the CPU 160 normalizes the coefficient values F01 to F07 of the first coefficient group based on the following equation (2) (step S330). According to the following equation (2), the normalized coefficient values Fr01 to Fr07 are values obtained by dividing the coefficient values F01 to F07 by the sum S of absolute values of the first coefficient group.

Fr0j = F0j / S (j = 1-7) (2)

  Next, the CPU 160 refers to the edge pattern table 181 stored in the ROM 180 (step S340), and the gradient shape represented by the normalized coefficient values Fr01 to Fr07 approximates any one of the basic edge patterns. Is determined (step S350).

  FIG. 14 is an explanatory diagram showing an example of the edge pattern table 181. In the edge pattern table 181, 16 basic edge patterns (third column in the figure) are recorded in association with pattern numbers (first column in the figure) from “A1” to “A16”. . The horizontal axis of the basic edge pattern represents the coefficient positions (F01 to F07) in the block, and the vertical axis represents the normalized coefficient value Fr. That is, the basic edge pattern is data composed of seven coefficient values.

  Each basic edge pattern is generated based on the luminance pattern shown in the second column of the figure. That is, the basic edge pattern shown in the third column of the figure is obtained by DCT transforming and normalizing the luminance pattern shown in the second column of the figure. Note that the luminance pattern in the figure is not actually recorded in the edge pattern table 181 but is described for convenience of understanding. Each luminance pattern is obtained by classifying typical shapes of luminance changes in which the inclination sign does not change within one block into 16 types in advance. In this embodiment, the basic edge patterns are classified into 16 types as described above, but may be classified into more patterns.

  The edge pattern table 181 further records three types of parameters associated with the basic edge pattern: a left edge width LW, a central edge width MW, and a right edge width RW. The left edge width LW represents the width of the flat portion present on the left side of the luminance pattern, and the right edge width RW represents the width of the flat portion present on the right side of the luminance pattern. Further, the center edge width MW represents the width of the gradient portion sandwiched between the left edge width LW and the right edge width RW.

  Although not shown, the edge pattern table 181 includes, in addition to the basic edge patterns “A1” to “A16”, edge patterns obtained by horizontally inverting these basic edge patterns having pattern numbers B1 to B1. The edge pattern defined in B16 and inverted up and down is defined in pattern numbers C1 to C16. Further, edge numbers obtained by inverting the basic edge patterns “A1” to “A16” vertically and horizontally are defined as pattern numbers D1 to D16. That is, 64 types of basic edge patterns are defined in the edge pattern table 181 as a whole.

  In step S350 in FIG. 13, the sum SD of the absolute values of the differences between the coefficient values Fr01 to Fr07 normalized in step S330 and the coefficient values Fb01 to Fb07 constituting the basic edge pattern is expressed by the following equation (3). The basic edge pattern having the minimum total value SD is selected from the edge pattern table 181. If the total value SD for the selected basic edge pattern is smaller than a predetermined threshold value, it is determined that a basic edge pattern whose shape approximates has been searched (step S350: Yes), and corresponds to the basic edge pattern. The attached pattern number is acquired (step S360). On the other hand, if the value SD obtained by the above-described calculation is larger than the predetermined threshold value, the pattern number is set to “unknown” because no approximate basic edge pattern has been searched (step S370).

SD = Σ | Fr0k−Fb0k | (k = 1 to 7) (3)

  When the pattern number of the current block is determined by the above processing, the CPU 160 next refers to the normalized luminance difference table 182 recorded in the ROM 180 (step S380), and calculates the luminance difference DY in the current block. (Step S390).

  FIG. 15 is an explanatory diagram illustrating an example of the normalized luminance difference table 182. In the normalized luminance difference table 182, a normalized luminance difference value NY is recorded in association with each pattern number. The normalized luminance difference value NY is a luminance difference calculated in advance according to the shape of the luminance pattern represented by the basic edge pattern. In step S390, the normalized luminance difference value NY associated with the pattern number acquired in step S360 is acquired from the normalized luminance difference table 182. Then, the luminance difference DY in the block is obtained by multiplying the acquired normalized luminance difference value NY by the sum S of absolute values of the respective coefficients calculated in the above equation (1) according to the following equation (4). Ask. As described above, if the normalized luminance difference value NY calculated in advance is used, the luminance difference in the block can be calculated easily and quickly without obtaining the luminance value by performing inverse DCT conversion on the coefficient value in the block. it can.

DY = S · NY (4)

  When the pattern number of the current block and the luminance difference DY are determined by the above processing, the CPU 160 stores these parameters in the pattern number buffer 171 (see FIG. 10 or FIG. 11) corresponding to the current block (see FIG. 10 or FIG. 11). Step S400). As a result, the above-described series of edge pattern matching processing ends, and the processing is returned to step S110 of the blur determination processing.

F. Edge connection processing:
FIG. 16 is a detailed flowchart of the edge connection process executed in step S150 of the blur determination process shown in FIG. As shown in FIG. 6, this process is a process for determining the blur width and the luminance difference of the central block (target block) based on blocks that are adjacent vertically and horizontally. Such processing is performed in the horizontal direction and the vertical direction, but in order to simplify the description, the processing performed in the horizontal direction will be described unless otherwise specified.

  When this process is executed, the CPU 160 first obtains the pattern number of the current block of interest and the luminance difference DY from the pattern number buffer 171 (see FIG. 10 or FIG. 11) (step S500). These parameters are parameters recorded in the pattern number buffer 171 by the edge pattern matching process.

  Next, CPU 160 determines whether or not the pattern number acquired in step S500 is “unknown” (step S520). As a result, if the pattern number is “unknown” (step S510: Yes), it is determined in the above-described edge pattern matching process that the luminance change of the current block of interest does not approximate any of the basic edge patterns. Therefore (see steps S350 and S370 in FIG. 13), the blur width of the block of interest is determined as “unknown” (step S520). Hereinafter, the determined blur width is referred to as a “determined blur width DW”.

  If it is determined in step S510 that the pattern number is not “unknown” (step S510: No), then the CPU 160 determines whether the pattern number is “0” (step S530). If the pattern number is “0” as a result of this determination (step S530: Yes), it is determined that the luminance change in the horizontal direction of the current block of interest is flat in the edge pattern matching process described above. Therefore (see steps S310 and S320 in FIG. 13), the determined blur width DW of the block of interest is determined to be “0” (step S550).

  If it is determined in step S530 that the pattern number is not “0” (step S530: No), the CPU 160 blocks adjacent blocks on the right side of the target block (lower side in the case of vertical edge connection processing). The pattern number (hereinafter referred to as “adjacent block”) is acquired from the pattern number buffer 171 (see FIG. 10 or FIG. 11) (step S510). Then, by comparing the pattern number of the target block with the pattern number of the adjacent block, it is determined whether the inclination directions of the basic edge patterns of these blocks match (step S560).

  FIG. 17 is an explanatory diagram illustrating an example in which the inclination directions of the basic edge patterns match between the target block and the adjacent block. In the example shown in the figure, since the slopes of both blocks are downwardly inclined to the right, it can be said that the slope directions of both blocks coincide. As combinations in which the inclination directions of the target block and the adjacent blocks match, there are the following pattern number combinations. That is, if the combination of the pattern number of the target block and the pattern number of the adjacent block is included in any one of the combinations (1) to (8) shown below, the CPU 160 determines that these inclination directions match. Can do.

Attention block No. Adjacent block No.
(1) A1-A16 A1-A16
(2) A1-A16 D1-D16
(3) B1-B16 B1-B16
(4) B1-B16 C1-C16
(5) C1-C16 C1-C16
(6) C1-C16 B1-B16
(7) D1 to D16 D1 to D16
(8) D1 to D16 A1 to A16

  FIG. 18 is an explanatory diagram illustrating an example in which the inclination directions of the basic edge pattern do not match between the target block and the adjacent block. In the illustrated example, the inclination of the block of interest is lower right, whereas the inclination of the adjacent block is upper right, it can be said that the inclinations of both do not match. Combinations of pattern numbers in which the inclination directions of both blocks do not match are combinations other than (1) to (8) described above.

  If it is determined in step S560 that the inclination directions are the same (step S560: Yes), the CPU 160 continues to add the right edge width RW of the target block and the left edge width LW of the adjacent block. It is determined whether it is within a predetermined connection error threshold (step S570). Prior to this determination, the CPU 160 obtains the left edge width LW with reference to the edge pattern table 181 from the pattern number of the adjacent block.

  FIG. 19 is an explanatory diagram showing a state of comparison between the added value of the right edge width RW of the target block and the left edge width LW of the adjacent block and the connection error threshold. As shown in the drawing, the sum of the flat portion (right edge width RW) present on the right side of the basic edge pattern of the target block and the flat portion (left edge width LW) present on the left side of the basic edge pattern of the adjacent block is large. When the connection error threshold is exceeded, it can be determined that the gradient is not continuous between the target block and the adjacent block, and separate blurred portions are formed. Therefore, it is determined as “No” in step S570. On the other hand, when the above-described sum is small and is equal to or less than the connection error threshold, it can be considered that the edge pattern of the block of interest and the edge pattern of the adjacent block have a continuous gradient. , “Yes”.

  If it is determined in step S570 that the added value of the right edge width RW of the target block and the left edge width LW of the adjacent block is within a predetermined connection error threshold (step S570: Yes), the CPU 160 The accumulated blur width CW and the accumulated luminance difference ADY accumulated in the block on the left side (upper in the case of vertical edge connection processing) are converted into the horizontal accumulation buffer 174 (in the vertical edge connection processing, Read from the vertical accumulation buffer 173) (see FIGS. 10 and 11). Then, the left edge width LW and the central edge width MW of the block of interest are added to the accumulated blur width CW, and the luminance difference DY of the block of interest is added to the accumulated luminance difference ADY (step S580). When the new cumulative blur width CW and the cumulative luminance difference ADY are obtained in this way, the CPU 160 overwrites and updates the new value in the horizontal cumulative buffer 174 (vertical cumulative buffer 173 in the case of vertical edge connection processing). (See FIGS. 10 and 11). In this way, when the inclination of the edge pattern between adjacent blocks continues in one direction, the blur width and luminance difference of each block can be accumulated sequentially. In step S580, CPU 160 determines the determined blur width DW of the block of interest as “0” for convenience. That is, when the blur width is accumulated, the block in the middle of accumulation is regarded as a flat block.

  FIG. 20 is an explanatory diagram showing the concept of calculating the cumulative blur width CW in step S580. As shown in the drawing, in step S580, a value obtained by adding the left edge width LW and the central edge width MW of the block of interest to the cumulative blur width CW so far is set as a new cumulative blur width CW. Prior to this calculation, the CPU 160 refers to the edge pattern table 181 to obtain the left edge width LW and the central edge width MW corresponding to the pattern number of the block of interest.

  When it is determined in step S560 that the inclination direction of the target block does not match the inclination direction of the adjacent block (step S560: No), or in step S570, the right edge width RW of the target block and the adjacent block When it is determined that the sum with the left edge width LW exceeds the connection error threshold (step S570: No), the CPU 160 subsequently determines whether or not the target block is flat using the accumulated luminance difference ADY. The process to judge is performed. That is, first, the accumulated luminance difference ADY is read from the horizontal accumulation buffer 174 (or the vertical accumulation buffer 173 in the case of vertical edge connection processing), and the sum of the accumulated luminance difference ADY and the luminance difference DY of the target block is calculated. Obtained (step S590). Then, it is determined whether this sum is less than a predetermined cumulative luminance difference threshold (for example, “32”) (step S600). If it is determined by this determination that the accumulated luminance difference ADY is less than the predetermined accumulated luminance difference threshold (step S600: Yes), the accumulated value is not sufficient even though the luminance difference DY has been accumulated so far. Then, it is determined that the luminance change of the block of interest is flat, and the determined blur width DW is set to “0” (step S610). According to such processing, it is possible to determine whether or not the block is flat not only based on the determination based on the edge pattern but also based on the luminance difference. Is possible.

  If it is determined in step S600 that the accumulated luminance difference ADY exceeds the accumulated luminance difference threshold (step S600: No), it can be determined that the block of interest is not a flat edge pattern. Therefore, the CPU 160 subsequently determines whether or not the accumulated blur width CW read from the horizontal accumulation buffer 174 (vertical accumulation buffer 173 in the case of vertical edge connection processing) is “0” (step S620). . If it is determined by this determination that the cumulative blur width CW is “0” (step S620: Yes), the block of interest is a starting point at which blurring occurs, and further, according to the determination results in steps S560 and S570, Since the edge pattern is not continuous with the adjacent block on the right side (the lower side in the case of edge connection processing in the vertical direction), the target block can be determined to have a single edge pattern. The CPU 160 determines the central edge width MW of the block of interest as the definite blur width DW (step S630) Note that the CPU 160 refers to the edge pattern table 181 prior to such processing and corresponds to the pattern number of the block of interest. Assume that the center edge width MW is acquired.

  If it is determined in step S620 that the accumulated blur width CW is not “0”, the target block is not continuous with the adjacent block on the right side (lower side in the edge connection processing in the vertical direction). The adjacent block on the left side (upper side in the case of edge connection processing in the vertical direction) is continuous. That is, the target block corresponds to the end of the blurred portion. Therefore, the CPU 160 adds the left edge width LW and the center edge width MW of the block of interest to the accumulated blur width CW read from the horizontal accumulation buffer 174 (vertical accumulation buffer 173 in the case of vertical edge linking processing). This value is determined as the definite blur width DW (step S640). Prior to this processing, the CPU 160 refers to the edge pattern table 181 and acquires the left edge width LW and the central edge width MW corresponding to the pattern number of the block of interest.

  When the determined blur width DW is determined in the above step S640, since the blur width is terminated, the CPU 160 subsequently sets (resets) the cumulative blur width CW and the cumulative luminance difference ADY to “0”. Is recorded in the horizontal accumulation buffer 174 (vertical accumulation buffer 173 in the case of vertical edge connection processing) (step S650). By doing so, the accumulation of the blur width is started from the position of the next block of interest. Even when the determined blur width DW is determined in steps S520, S540, S610, and S630, the cumulative blur width CW and the cumulative luminance difference ADY are set to “0” in step S650. In these cases, the block of interest is flat or forms a single edge, so there is no need to accumulate blur width and luminance difference.

  On the other hand, if it is determined in step S570 that the target block and the adjacent block are continuous, the cumulative blur width CW and the cumulative luminance difference ADY are not reset, and the cumulative value calculated in step S580 is laterally cumulative. The data is recorded in the buffer 174 (vertical accumulation buffer 173 in the case of vertical edge connection processing). This makes it possible to accumulate the blur width and the luminance difference for the next block of interest.

  This completes the series of edge connection processing. According to this edge connection process, the horizontal and vertical definite blur widths DW used in the block blur determination process described later are determined, and the cumulative blur width CW and the cumulative luminance difference ADY used in the next edge connection process are determined. Are recorded in the horizontal accumulation buffer 174 and the vertical accumulation buffer 173. When the edge connection process is completed, the CPU 160 returns the process to the blur determination process shown in FIG. 9, and subsequently executes the block blur determination process (step S160).

G. Block blur judgment processing:
FIG. 21 is a detailed flowchart of the block blur determination process executed in step S160 of the blur determination process shown in FIG. This process is a process executed subsequent to the above-described edge connection process, and is a process for determining whether or not the block of interest is in focus.

  When this process is executed, the CPU 160 first acquires the determined blur width DW in the horizontal and vertical directions of the block of interest determined in the edge connection process (step S700).

  When the two determined blur widths DW are acquired, the CPU 160 determines whether one of these determined blur widths DW is “unknown” (step S710). If it is determined by this determination that one of the determined blur widths DW is “unknown” (step S710: Yes), the edge pattern of the block of interest does not approximate any of the basic edge patterns. The block is uniformly determined to be a “blurred block” (step S720).

  On the other hand, if it is determined in step S710 that neither of the determined blur widths DW is “unknown” (step S710: No), the CPU 160 determines the larger one of the two determined blur widths DW. The blur width DW is specified as the maximum determined blur width MDW (step S730).

  When the maximum determined blur width MDW is specified, the CPU 160 determines whether or not the maximum determined blur width MDW is larger than “0” and smaller than a predetermined blur width threshold (for example, “16”) ( Step S740). If the maximum definite blur width MDW satisfies this condition as a result of the determination, the target block is determined to be a “focus block” (step S750). On the other hand, if the maximum defined blur width MDW is “0” or is equal to or greater than a predetermined blur width threshold, it is determined that the block of interest is a “blurred block” (step S720).

  When the determination of whether the current block of interest is “in-focus block” or “blurred block” is completed by the above processing, the determination result is a sub-window (FIG. 7) in step S170 of the blur determination process shown in FIG. (Refer to each). The number of in-focus blocks thus counted is stored in the in-focus block number buffer 172 of the RAM 170.

  According to the block blur determination process described above, the determined blur width DW having the larger value of the horizontal determined blur width DW and the vertical determined blur width DW is specified as the maximum determined blur width MDW. If the maximum determined blur width MDW is larger than the predetermined blur width threshold, it can be determined that the block of interest is blurred.

H. Window blur determination processing:
FIG. 22 is a detailed flowchart of the window blur determination process executed in step S190 of the blur determination process shown in FIG. The above-described block blur determination process is a process for determining the presence / absence of blur for each block, whereas this window blur determination process is a process for determining the presence / absence of blur for each window area shown in FIG. This process is executed when the block blur determination process for all the blocks belonging to one subwindow is completed.

  When this process is executed, the CPU 160 first refers to the focused block number buffer 172 shown in FIG. 12 (step S800). Then, the cumulative value of the number of in-focus blocks is acquired for all sub-windows in the window area currently focused on (hereinafter referred to as “target window”) (step S810), and these are added together to obtain the window area. The number of all the in-focus blocks is tabulated (step S820).

  FIG. 23 is an explanatory diagram showing a window area moving method and a method of counting the number of focused blocks. FIG. 23A shows the initial position of the window of interest, and the cumulative value of the number of focused blocks in the sub-window at the upper left corner of the image is “3” in accordance with the recording order of the number of focused blocks shown in FIG. Is recorded. In this case, in step S820, all the in-focus block numbers (0, 0, 0, 3) shown in the first column of the figure are added. If 4 × 4 subwindows are included in one window area and 7 × 7 blocks are included in one subwindow, if all the blocks in the window area are in focus, then 1 The total number of focused blocks for the window area is 784 (= 7 × 7 × 16).

  Returning to FIG. When the number of in-focus blocks is counted in step S820, the CPU 160 determines whether the total number of in-focus blocks is equal to or greater than a predetermined focus density threshold (for example, 400) (step S830). If the result of this determination is that the total number of in-focus blocks is equal to or greater than the focus density threshold (step S830: Yes), the current window of interest is determined to be the “focus window” (step S840), and the focus density threshold is met. If not (Step S830: No), it is determined as a “blurred window” (Step S850).

  If it is determined in step S840 that the window is “focused”, the CPU 160 ends the window blur determination process. Then, the process proceeds to step S210 of the blur determination process of FIG. 9, and it is determined that the entire image is a focused image. On the other hand, if it is determined in step S850 that the window is a blurred window, the window of interest is moved to the right by one column in the focused block number buffer 172 (step S860). FIG. 22B shows an example in which the window is moved by one column from FIG. Thereafter, with this process, the window of interest gradually moves to the right as shown in FIGS. By doing so, the number of in-focus windows counted in step S180 of the blur determination process in FIG. 9 is recorded in the lower right corner of the window area.

  When the window of interest is moved in step S860, the CPU 160 determines whether the window area has moved to the right end as shown in FIG. 22E (step S870). If it is determined by this processing that the engine has not been started to the right end (step S870: No), CPU 160 ends the window blur determination process and returns the process to step S200 of the blur determination process shown in FIG.

  On the other hand, if it is determined that the window of interest has moved to the right end (step S870: Yes), a process of moving up the row of the focused block number buffer 172 is performed (step S880). This carry-up process is a process for raising the data in the fourth row of the focused block number buffer to the third row, the third row to the second row, and the second row to the first row. At this time, for the first row, the row is moved down to the fourth row, and at the same time, all data in this row are initialized. FIG. 22 (e) shows the state of the carry processing. As shown in FIGS. 3 and 4, reading of blocks from JPEG data basically starts from the upper left of the image, whereas in this process, as shown in FIG. The data is filled in order from the bottom row. In this carry-up process, the data stored in the RAM 170 may be moved for each row, but the address pointers that refer to each row may be sequentially replaced. If the pointer is used in this way, it is not necessary to move data in the RAM, so that the processing time can be shortened.

  After moving up the row of the focused block number buffer 172 in step S880, the CPU 160 moves the window of interest to the left end of the focused block number buffer 172 as shown in FIG. 22 (f) (step S890). . Then, the next focused block number is recorded from the first column of the last row of the focused block number buffer 172. If it is determined that the target window is a “blurred window”, the process from step S860 to step S890 described above is executed, so that the target window is moved in units of one subwindow from the upper left to the lower right in the image. Can be made.

I. effect:
According to the printer 100 of the present embodiment described above, the presence or absence of blur is determined when the JPEG data is expanded to the DCT coefficient that is the frequency domain without being expanded to the spatial domain represented by RGB, YCbCr, or the like. can do. Therefore, the blur determination process can be performed at high speed. Further, in this embodiment, the presence / absence of blur is determined only by a total of 14 DCT coefficients constituting the first coefficient group and the second coefficient group shown in FIG. 5 among the DCT coefficients of 8 pixels × 8 pixels. Therefore, the amount of calculation can be reduced.

  Further, in this embodiment, the first coefficient group and the second coefficient group are not stored in the RAM 170 as they are, but the basic edge pattern approximated to the luminance change represented by these coefficient groups is obtained from the edge pattern table 181. The search is performed and the pattern number is stored in the pattern number buffer 171 shown in FIGS. Therefore, the memory capacity can be greatly reduced. Also, as for the area for storing the pattern number, as shown in FIG. 10 and FIG. 11, only the area where the blocks adjacent to the right side and the lower side of the target block for which the presence or absence of blur is to be recorded can be recorded is prepared. As a result, the memory capacity to be used can be significantly reduced.

  In this embodiment, the presence or absence of blurring of the entire image is determined based on a predetermined window area that finally includes a plurality of blocks. The focused block number buffer 172 used at this time also includes a sub-window included in the window area. It is assumed that only the number of rows is prepared. Therefore, the memory capacity to be used can be reduced.

  Further, in this embodiment, the blur widths of adjacent blocks can be sequentially connected by the edge connection process described above, and therefore, the width of the blurred portion in the image exceeds the size of the 8 × 8 unit block. Also, it is possible to determine whether or not the image is blurred with high accuracy.

  As mentioned above, although the Example of this invention was described, it cannot be overemphasized that this invention is not limited to such an Example, and can take a various structure in the range which does not deviate from the meaning.

  For example, in the above embodiment, the printer 100 performs the blur determination. However, the computer may perform the blur determination by installing a program for executing the above-described various processes in the computer.

  Moreover, although the case where the size of one block is 8 × 8 has been described in the above embodiment, the block size is not limited to this. Further, the size of the window area and the number of subwindows included in one window area can be set as appropriate.

  In the above embodiment, the number of blocks read from the image data is minimized by sequentially moving the block of interest. However, the edge pattern matching process is performed on all the blocks included in the image data. Then, the pattern number may be stored, and the edge connection process or the block blur determination process may be performed using the pattern number of the entire image thus stored. This also reduces the amount of memory used because it is not necessary to hold the coefficient values of the entire image.

1 is an explanatory diagram illustrating an appearance of a printer 100 as an embodiment. FIG. 2 is an explanatory diagram illustrating an internal configuration of a printer 100. FIG. It is explanatory drawing which shows a mode that each block is recorded for every line in JPEG data. It is explanatory drawing which shows a mode that each block is recorded for every 2 lines in JPEG data. It is explanatory drawing which shows the DCT coefficient extracted by the coefficient extraction part 162. FIG. It is explanatory drawing which shows the concept which connects the blur width of an attention block and the block adjacent to this. It is explanatory drawing which shows the concept of a window area | region. It is a flowchart of a printing process. It is a flowchart of a blurring determination process. It is explanatory drawing which shows an example of the pattern number buffer 171 ensured when the block is recorded by JPEG data per line. It is explanatory drawing which shows an example of the pattern number buffer 171 ensured when the block is recorded by JPEG data per 2 lines. 5 is an explanatory diagram illustrating an example of a focused block number buffer 172. FIG. It is a detailed flowchart of an edge pattern collation process. 5 is an explanatory diagram illustrating an example of an edge pattern table 181. FIG. It is explanatory drawing which shows an example of the normalization brightness | luminance difference table. It is a detailed flowchart of an edge connection process. It is explanatory drawing which shows the example in which the inclination direction of a basic edge pattern corresponds with an attention block and an adjacent block. It is explanatory drawing which shows the example from which the inclination direction of a basic edge pattern does not correspond with an attention block and an adjacent block. It is explanatory drawing which shows the mode of a comparison with the addition value of the right edge width RW of an attention block, and the left edge width LW of an adjacent block, and a connection error threshold value. It is explanatory drawing which shows the calculation concept of the accumulation blur width CW. It is a detailed flowchart of a block blur determination process. It is a detailed flowchart of a window blurring determination process. It is explanatory drawing which shows the movement method of a window area | region, and the total method of the number of focused blocks.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Printer 110 ... Scanner 120 ... Memory card slot 130 ... USB interface 140 ... Operation panel 145 ... Liquid crystal display 150 ... Control unit 160 ... CPU
161 ... Image data input unit 162 ... Coefficient extraction unit 163 ... Pattern matching unit 164 ... Edge connection unit 165 ... Block blur determination unit 166 ... Window blur determination unit 170 ... RAM
171 ... Pattern number buffer 172 ... Focus block number buffer 173 ... Vertical accumulation buffer 174 ... Horizontal accumulation buffer 180 ... ROM
181 ... Edge pattern table 182 ... Normalized luminance difference table 210 ... Carriage 211 ... Ink head 212 ... Ink cartridge 220 ... Carriage motor 230 ... Paper feed motor 260 ... Drive belt 270 ... Platen 280 ... Sliding shaft

Claims (17)

An image processing apparatus for detecting blur of an image configured as a set of blocks composed of a plurality of pixels,
An image data input unit for inputting image data in which coefficients obtained by conversion from a spatial domain to a frequency domain performed in units of the blocks are recorded for each block; and
A plurality of types of basic edge patterns in which representative gradient shapes of pixel value changes in the block are represented by the coefficients, and pattern numbers uniquely assigned to the basic edge patterns are stored in association with each other. An edge pattern storage unit;
A coefficient extraction unit that extracts a coefficient group representing a frequency component in a predetermined direction from each block of the input image data;
A pattern matching unit that selects a basic edge pattern that approximates a gradient shape represented by the extracted coefficient group from the edge pattern storage unit;
A pattern number storage unit that stores the pattern number corresponding to the selected basic edge pattern in association with the block;
A blur width detecting unit that refers to a pattern number stored in the pattern number storage unit and detects a range in which the gradient directions of the basic edge pattern coincide with each other in a series of blocks in the image as a blur width;
An image processing apparatus comprising: a blur determination unit that determines blur of an image represented by the image data based on the detected blur width. The image processing apparatus according to claim 1,
The edge pattern storage unit further stores a gradient width represented by each basic edge pattern in association with the pattern number,
The blur width detection unit reads the gradient width associated with the pattern number of each block existing in a range where the gradient directions of the basic edge pattern coincide with each other from the edge pattern storage unit, and calculates the gradient width. An image processing apparatus that calculates the blur width by adding. The image processing apparatus according to claim 2,
The blur width detection unit moves a block of interest in the image data along the series of blocks, and is associated with the block of interest and an adjacent block to which the block of interest is moved, respectively. Referring to the pattern number, it is determined whether or not the direction of the gradient of the basic edge pattern of the block of interest and the adjacent block matches, and if the direction of the gradient matches, it corresponds to the pattern number of the block of interest An image processing apparatus that calculates the blur width by accumulatively adding the attached gradient widths. The image processing apparatus according to claim 3,
In the image data, the blocks are arranged in a horizontal direction and a vertical direction,
The coefficient extraction unit extracts, as the coefficient group, a first coefficient group representing a horizontal frequency component and a second coefficient group representing a vertical frequency component,
The pattern matching unit selects the basic edge pattern for each of the first coefficient group and the second coefficient group,
The pattern number storage unit stores the pattern numbers corresponding to the selected two basic edge patterns in association with the blocks, respectively.
The blur width detection unit calculates the blur width in the horizontal direction and the vertical direction,
The blur determination unit determines blur of an image represented by the image data based on a larger blur width among the blur widths calculated in the horizontal direction and the vertical direction, respectively. The image processing apparatus according to claim 4,
The blur width detecting unit stores the gradient width cumulatively added in the horizontal direction in a horizontal cumulative buffer that holds one block, and holds the gradient width cumulatively added in the vertical direction for one block. An image processing apparatus that stores in a vertical accumulation buffer and acquires the horizontal and vertical gradient widths accumulated in a block before movement by referring to these buffers when the block of interest is moved. The image processing apparatus according to claim 4 or 5, wherein
The pattern number storage unit has a storage area capable of storing the pattern number for a first row including the target block and a second row including a block adjacent to the lower side of the target block, When the block of interest moves to the right end of the row to which the block of interest belongs, the pattern number stored in the second and subsequent rows is moved up to the previous row and stored, so that the next selected block The image number is stored in the empty area of the pattern number storage unit. The image processing apparatus according to claim 6,
In the image data, the block is recorded every n (n is an integer of 1 or more) rows,
The pattern processing unit prepares a storage area capable of storing pattern numbers for the block (n + 1) rows. An image processing apparatus according to any one of claims 1 to 7,
The blur determination unit
A block blur determination unit that determines the presence or absence of blur for each block based on the blur width and a predetermined threshold;
For all blocks included in the predetermined window area in the image data, the presence / absence of blur by the block blur determination means is totaled, and whether or not the window area is blurred based on the total result and a predetermined threshold value A window blur determination unit for determining
As a result of moving the window area in the image data, if there is at least one window area that is determined not to be blurred, it is determined that the entire image represented by the image data is not blurred. An image processing apparatus comprising: an image blur determination unit. The image processing apparatus according to claim 8,
The window area is further partitioned by a plurality of sub-windows, each of which is larger than the size of the block,
The block blur determination unit totalizes the presence / absence of blur for each block for each sub-window including the block,
The window blur determination unit further counts the total value of the sub-windows, and further totals all the sub-windows included in the window area, thereby calculating the presence / absence of the blur for all blocks included in the window area. Perform image processing device. The image processing apparatus according to claim 9,
The image blur determination unit is an image processing apparatus that moves the window area in the image data by a movement distance in units of sub-windows. The image processing apparatus according to claim 10,
Furthermore, a total value storage unit that stores a total value of the presence or absence of blur for each of the sub-windows totaled by the block blur determination unit,
The total value storage unit has a storage area for storing the total value as many as the number of rows of subwindows included in the window area, and when the window area is moved downward in the image data The total value stored in each row of the total value storage unit is stored in the previous row by the number of the moved lines, and the empty region of the total value storage unit generated by the advance processing An image processing apparatus for storing the total value for each of the sub-windows that are included in the window area after the movement. An image processing apparatus according to any one of claims 1 to 11,
Furthermore, an image processing apparatus provided with the presentation part which shows a user the image determined not to be blurred by the blur determination part. The image processing apparatus according to claim 12,
An image processing apparatus further comprising a printing unit that prints an image selected by the user from the presented images. An image processing apparatus according to any one of claims 1 to 13,
The image processing apparatus is an image processing device that is image data in JPEG format. A blur detection method in which a computer detects blur of an image configured as a set of blocks made up of a plurality of pixels,
Coefficients obtained by conversion from the spatial domain to the frequency domain performed in units of the blocks are input image data recorded and configured for each block,
Extracting a coefficient group representing a frequency component in a predetermined direction from each block of the input image data,
A plurality of types of basic edge patterns in which representative gradient shapes of pixel value changes in the block are represented by the coefficients, and pattern numbers uniquely assigned to the basic edge patterns are stored in association with each other. With reference to the edge pattern storage unit, a basic edge pattern that approximates the gradient shape represented by the extracted coefficient group is selected,
Storing the pattern number corresponding to the selected basic edge pattern in a predetermined pattern number storage unit in association with the block;
With reference to the pattern number stored in the pattern number storage unit, a range in which the direction of the gradient of the basic edge pattern matches along the series of blocks in the image is detected as a blur width,
A blur detection method for determining blur of an image represented by the image data based on the detected blur width. A computer program for detecting blurring of an image configured as a set of blocks composed of a plurality of pixels,
A function of inputting image data in which a coefficient obtained by conversion from a spatial domain to a frequency domain performed in units of the block is recorded for each block;
A function of extracting a coefficient group representing a frequency component in a predetermined direction from each block of the input image data;
A plurality of types of basic edge patterns in which representative gradient shapes of pixel value changes in the block are represented by the coefficients, and pattern numbers uniquely assigned to the basic edge patterns are stored in association with each other. A function of selecting a basic edge pattern that approximates a gradient shape represented by the extracted coefficient group with reference to an edge pattern storage unit;
A function of storing the pattern number corresponding to the selected basic edge pattern in the predetermined pattern number storage unit in association with the block;
A function of referring to a pattern number stored in the pattern number storage unit and detecting as a blur width a range in which the direction of the gradient of the basic edge pattern coincides along the series of blocks in the image;
A computer program that causes a computer to realize a function of determining blur of an image represented by the image data based on the detected blur width.   The computer-readable recording medium which recorded the computer program of Claim 16.

Images